d10b76552f
The data-layout independent constant folding currently has some rather gnarly code for canonicalizing GEP indices to reduce "notional overindexing", and then infers inbounds based on that canonicalization. Now that we canonicalize to i8 GEPs, this canonicalization is essentially useless, as we'll discard it as soon as the GEP hits the data-layout aware constant folder anyway. As such, I'd like to remove this code entirely. This shouldn't have any impact on optimization capabilities.
108 lines
5.6 KiB
LLVM
108 lines
5.6 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=infer-address-spaces %s | FileCheck %s
|
|
|
|
; Test that pure GetElementPtr instructions not directly connected to
|
|
; a memory operation are inferred.
|
|
|
|
@lds = internal unnamed_addr addrspace(3) global [648 x double] undef, align 8
|
|
|
|
define void @simplified_constexpr_gep_addrspacecast(i64 %idx0, i64 %idx1) {
|
|
; CHECK-LABEL: @simplified_constexpr_gep_addrspacecast(
|
|
; CHECK-NEXT: [[GEP0:%.*]] = getelementptr inbounds double, ptr addrspace(3) getelementptr inbounds ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384), i64 [[IDX0:%.*]]
|
|
; CHECK-NEXT: store double 1.000000e+00, ptr addrspace(3) [[GEP0]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%gep0 = getelementptr inbounds double, ptr addrspacecast (ptr addrspace(3) getelementptr inbounds ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384) to ptr), i64 %idx0
|
|
%asc = addrspacecast ptr %gep0 to ptr addrspace(3)
|
|
store double 1.000000e+00, ptr addrspace(3) %asc, align 8
|
|
ret void
|
|
}
|
|
|
|
define void @constexpr_gep_addrspacecast(i64 %idx0, i64 %idx1) {
|
|
; CHECK-LABEL: @constexpr_gep_addrspacecast(
|
|
; CHECK-NEXT: [[GEP0:%.*]] = getelementptr inbounds double, ptr addrspace(3) getelementptr ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384), i64 [[IDX0:%.*]]
|
|
; CHECK-NEXT: store double 1.000000e+00, ptr addrspace(3) [[GEP0]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%gep0 = getelementptr inbounds double, ptr getelementptr ([648 x double], ptr addrspacecast (ptr addrspace(3) @lds to ptr), i64 0, i64 384), i64 %idx0
|
|
%asc = addrspacecast ptr %gep0 to ptr addrspace(3)
|
|
store double 1.0, ptr addrspace(3) %asc, align 8
|
|
ret void
|
|
}
|
|
|
|
define void @constexpr_gep_gep_addrspacecast(i64 %idx0, i64 %idx1) {
|
|
; CHECK-LABEL: @constexpr_gep_gep_addrspacecast(
|
|
; CHECK-NEXT: [[GEP0:%.*]] = getelementptr inbounds double, ptr addrspace(3) getelementptr ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384), i64 [[IDX0:%.*]]
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds double, ptr addrspace(3) [[GEP0]], i64 [[IDX1:%.*]]
|
|
; CHECK-NEXT: store double 1.000000e+00, ptr addrspace(3) [[GEP1]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%gep0 = getelementptr inbounds double, ptr getelementptr ([648 x double], ptr addrspacecast (ptr addrspace(3) @lds to ptr), i64 0, i64 384), i64 %idx0
|
|
%gep1 = getelementptr inbounds double, ptr %gep0, i64 %idx1
|
|
%asc = addrspacecast ptr %gep1 to ptr addrspace(3)
|
|
store double 1.0, ptr addrspace(3) %asc, align 8
|
|
ret void
|
|
}
|
|
|
|
; Don't crash
|
|
define amdgpu_kernel void @vector_gep(<4 x ptr addrspace(3)> %array) nounwind {
|
|
; CHECK-LABEL: @vector_gep(
|
|
; CHECK-NEXT: [[CAST:%.*]] = addrspacecast <4 x ptr addrspace(3)> [[ARRAY:%.*]] to <4 x ptr>
|
|
; CHECK-NEXT: [[P:%.*]] = getelementptr [1024 x i32], <4 x ptr> [[CAST]], <4 x i16> zeroinitializer, <4 x i16> <i16 16, i16 16, i16 16, i16 16>
|
|
; CHECK-NEXT: [[P0:%.*]] = extractelement <4 x ptr> [[P]], i32 0
|
|
; CHECK-NEXT: [[P1:%.*]] = extractelement <4 x ptr> [[P]], i32 1
|
|
; CHECK-NEXT: [[P2:%.*]] = extractelement <4 x ptr> [[P]], i32 2
|
|
; CHECK-NEXT: [[P3:%.*]] = extractelement <4 x ptr> [[P]], i32 3
|
|
; CHECK-NEXT: store i32 99, ptr [[P0]], align 4
|
|
; CHECK-NEXT: store i32 99, ptr [[P1]], align 4
|
|
; CHECK-NEXT: store i32 99, ptr [[P2]], align 4
|
|
; CHECK-NEXT: store i32 99, ptr [[P3]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%cast = addrspacecast <4 x ptr addrspace(3)> %array to <4 x ptr>
|
|
%p = getelementptr [1024 x i32], <4 x ptr> %cast, <4 x i16> zeroinitializer, <4 x i16> <i16 16, i16 16, i16 16, i16 16>
|
|
%p0 = extractelement <4 x ptr> %p, i32 0
|
|
%p1 = extractelement <4 x ptr> %p, i32 1
|
|
%p2 = extractelement <4 x ptr> %p, i32 2
|
|
%p3 = extractelement <4 x ptr> %p, i32 3
|
|
store i32 99, ptr %p0
|
|
store i32 99, ptr %p1
|
|
store i32 99, ptr %p2
|
|
store i32 99, ptr %p3
|
|
ret void
|
|
}
|
|
|
|
define void @repeated_constexpr_gep_addrspacecast(i64 %idx0, i64 %idx1) {
|
|
; CHECK-LABEL: @repeated_constexpr_gep_addrspacecast(
|
|
; CHECK-NEXT: [[GEP0:%.*]] = getelementptr inbounds double, ptr addrspace(3) getelementptr ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384), i64 [[IDX0:%.*]]
|
|
; CHECK-NEXT: store double 1.000000e+00, ptr addrspace(3) [[GEP0]], align 8
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds double, ptr addrspace(3) getelementptr ([648 x double], ptr addrspace(3) @lds, i64 0, i64 384), i64 [[IDX1:%.*]]
|
|
; CHECK-NEXT: store double 1.000000e+00, ptr addrspace(3) [[GEP1]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%gep0 = getelementptr inbounds double, ptr getelementptr ([648 x double], ptr addrspacecast (ptr addrspace(3) @lds to ptr), i64 0, i64 384), i64 %idx0
|
|
%asc0 = addrspacecast ptr %gep0 to ptr addrspace(3)
|
|
store double 1.0, ptr addrspace(3) %asc0, align 8
|
|
|
|
%gep1 = getelementptr inbounds double, ptr getelementptr ([648 x double], ptr addrspacecast (ptr addrspace(3) @lds to ptr), i64 0, i64 384), i64 %idx1
|
|
%asc1 = addrspacecast ptr %gep1 to ptr addrspace(3)
|
|
store double 1.0, ptr addrspace(3) %asc1, align 8
|
|
|
|
ret void
|
|
}
|
|
|
|
define void @unorder_constexpr_gep_bitcast() {
|
|
; CHECK-LABEL: @unorder_constexpr_gep_bitcast(
|
|
; CHECK-NEXT: [[X0:%.*]] = load i32, ptr addrspace(3) @lds, align 4
|
|
; CHECK-NEXT: [[X1:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @lds, i32 1), align 4
|
|
; CHECK-NEXT: call void @use(i32 [[X0]], i32 [[X1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%x0 = load i32, ptr addrspacecast (ptr addrspace(3) @lds to ptr), align 4
|
|
%x1 = load i32, ptr getelementptr (i32, ptr addrspacecast (ptr addrspace(3) @lds to ptr), i32 1), align 4
|
|
call void @use(i32 %x0, i32 %x1)
|
|
ret void
|
|
}
|
|
|
|
declare void @use(i32, i32)
|